[INFRA] Import NVIDIA/CCCL upstream as optimization reference library

CCCL (CUDA C++ Core Libraries) provides:
- CUB: device/block/warp-level GPU primitives (reduce, scan, sort, topk)
- Thrust: high-level parallel algorithms (transform_reduce, sort, scan)
- libcudacxx: CUDA C++ standard library (atomics, barriers, memory)
- cudax: experimental features (memory resources, allocators)
- Tuning policies: per-SM hardware-specific algorithm parameters

Competition optimization vectors mapped to CCCL:
- Output TPS (83% weight): warp_reduce, block_reduce, device_topk
- Input TPS (14% weight): device_scan, block_load, prefetch
- Cache TPS (3% weight): prefix caching strategy patterns
- Memory (0.9 util): pooled/cached/buddy allocators

Source: https://github.com/NVIDIA/cccl (shallow clone, HEAD only)
License: Apache-2.0

cccl_upstream/libcudacxx/test/support/concurrent_agents.h

@@ -0,0 +1,52 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+// SPDX-FileCopyrightText: Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CONCURRENT_AGENTS_H
+#define _CONCURRENT_AGENTS_H
+
+#ifndef __CUDA_ARCH__
+#  include <thread>
+#endif
+
+#include <cuda/std/cassert>
+
+#include "test_macros.h"
+
+_CCCL_EXEC_CHECK_DISABLE
+template <class Fun>
+TEST_FUNC void execute_on_main_thread(Fun&& fun)
+{
+  NV_IF_ELSE_TARGET(NV_IS_DEVICE, (if (threadIdx.x == 0) { fun(); } __syncthreads();), (fun();))
+}
+
+template <typename... Fs>
+TEST_FUNC void concurrent_agents_launch(Fs... fs)
+{
+  NV_IF_ELSE_TARGET(
+    NV_IS_DEVICE,
+    (assert(blockDim.x == sizeof...(Fs)); using fptr = void (*)(void*);
+
+     fptr device_threads[] = {[](void* data) {
+       (*reinterpret_cast<Fs*>(data))();
+     }...};
+
+     void* device_thread_data[] = {reinterpret_cast<void*>(&fs)...};
+
+     __syncthreads();
+
+     device_threads[threadIdx.x](device_thread_data[threadIdx.x]);
+
+     __syncthreads();),
+    (std::thread threads[]{std::thread{std::forward<Fs>(fs)}...};
+
+     for (auto&& thread : threads) { thread.join(); }))
+}
+
+#endif // _CONCURRENT_AGENTS_H